On Wed, May 17, 2000 at 12:54:22PM +0200, Michael Schmitz wrote:
quoted
Tim Wojtulewicz wrote:
quoted
No it's definitely supposed to have a fan. It comes on in MacOS.
The problem is that the PMU is not fully supported yet, so it doesn't
understand that the processor is too hot. I'm trying to find
^^^^^^^^^^^^^^^^^^^^
Does this mean that I can damage the machine if I just run a high-load
program?!?
I strongly doubt that. The PMU should handle this autonomously, regardless
of Linux support for the PMU.
Also, i think the ppc cpu will halt itself when becoming too hot, and not let
itself burn. At elast it was so since the earlier 680x0 cpus.
Friendly,
Sven LUTHER
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
At 2:08 PM +0200 5/17/00, Sven LUTHER wrote:
On Wed, May 17, 2000 at 12:54:22PM +0200, Michael Schmitz wrote:
quoted
> Tim Wojtulewicz wrote:
> > No it's definitely supposed to have a fan. It comes on in MacOS.
> > The problem is that the PMU is not fully supported yet, so it doesn't
> > understand that the processor is too hot. I'm trying to find
> ^^^^^^^^^^^^^^^^^^^^
> Does this mean that I can damage the machine if I just run a high-load
> program?!?
I strongly doubt that. The PMU should handle this autonomously, regardless
of Linux support for the PMU.
Also, i think the ppc cpu will halt itself when becoming too hot,
No, it will not.
The PowerPC 750 (G3) and 7400 (G4) can both fire off an interrupt
when the on-die temperature sensor reading rises above a trigger
value (or falls below a second trigger value). This feature *could*
be used by an operating system to slow down the CPU (through the
instruction cache throttling feature) or halt it to prevent
overheating. However, there is no hardware feature which can halt
the CPU without software control.
and not let
itself burn. At elast it was so since the earlier 680x0 cpus.
As far as I know none of the 680x0 CPUs even had an on-die
temperature sensor, let alone a thermal shutdown feature.
I just talked with somebody who worked on the 101 project (popularly
and incorrectly known as "Lombard"). At least for that machine, the
fan is controlled as follows:
The normal mode is that MacOS controls the fan through the PMU. The
PMU is connected to a thermistor located somewhere on the
motherboard. MacOS scans the value of this thermistor every so often
and uses the fan to cool the machine if it's getting warm. MacOS
takes advantage of the speed control feature of the fan, so it won't
crank the fan up to full speed if it doesn't have to.
If MacOS fails to turn the fan on for some reason, the PMU acts as a
backup controller -- it is also monitoring the thermistor and will
force the fan to turn on at full speed if a moderately high (but not
yet unsafe) temperature is reached.
If the PMU's fan override fails to stop the computer's temperature
from rising, and the temperature begins to approach dangerous levels,
the PMU will simply shut the machine down without asking. (Most Macs
since the Mac II have had a thermal shutdown feature like this,
including the desktops.)
So, at least for 101, the fan should still come on without operating
system support, followed by an abrupt shutdown if the computer gets
way too hot.
The current PowerBook G3, 102 (the one with FireWire, popularly and
in this case correctly known as "Pismo"), does have a different PMU
than 101. Apple did a major upgrade to the PMU used with all Core99
chipset machines, 102 included. But it's fairly unlikely that Apple
changed the fan control technique very much.
Tim Seufert
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
On Wed, 17 May 2000, Timothy A. Seufert wrote:
No, it will not.
The PowerPC 750 (G3) and 7400 (G4) can both fire off an interrupt
when the on-die temperature sensor reading rises above a trigger
value (or falls below a second trigger value). This feature *could*
be used by an operating system to slow down the CPU (through the
instruction cache throttling feature) or halt it to prevent
overheating. However, there is no hardware feature which can halt
the CPU without software control.
So in other words, without a little bit of help from the OS, the fan won't
come on when it the temperature hits this trigger value. The temperature
would of course, continue to rise.
The normal mode is that MacOS controls the fan through the PMU. The
PMU is connected to a thermistor located somewhere on the
motherboard. MacOS scans the value of this thermistor every so often
and uses the fan to cool the machine if it's getting warm. MacOS
takes advantage of the speed control feature of the fan, so it won't
crank the fan up to full speed if it doesn't have to.
Has any code been written for LinuxPPC that will let us monitor this
sensor? I know there's a whole lm_sensors package for i386 and the like,
but is there something like this for us?
If MacOS fails to turn the fan on for some reason, the PMU acts as a
backup controller -- it is also monitoring the thermistor and will
force the fan to turn on at full speed if a moderately high (but not
yet unsafe) temperature is reached.
If the PMU's fan override fails to stop the computer's temperature
from rising, and the temperature begins to approach dangerous levels,
the PMU will simply shut the machine down without asking. (Most Macs
since the Mac II have had a thermal shutdown feature like this,
including the desktops.)
So, at least for 101, the fan should still come on without operating
system support, followed by an abrupt shutdown if the computer gets
way too hot.
Yes, my fan does come on Linux. But it takes a long time, and the case
gets very warm before it does. I assume this means the PMU has finally
decided it's too hot and turned on the fan. I know it's most likely not
Linux doing it.
The current PowerBook G3, 102 (the one with FireWire, popularly and
in this case correctly known as "Pismo"), does have a different PMU
than 101. Apple did a major upgrade to the PMU used with all Core99
chipset machines, 102 included. But it's fairly unlikely that Apple
changed the fan control technique very much.
Yes, I had heard they upgraded the PMU, but I wasn't aware of how much
they had changed it. Is it just a matter of knowing some extra bits we
can switch on and off in the PMU, or is it more than that?
Tim
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
At 7:48 AM -0700 5/18/00, Tim Wojtulewicz wrote:
On Wed, 17 May 2000, Timothy A. Seufert wrote:
quoted
No, it will not.
The PowerPC 750 (G3) and 7400 (G4) can both fire off an interrupt
when the on-die temperature sensor reading rises above a trigger
value (or falls below a second trigger value). This feature *could*
be used by an operating system to slow down the CPU (through the
instruction cache throttling feature) or halt it to prevent
overheating. However, there is no hardware feature which can halt
the CPU without software control.
So in other words, without a little bit of help from the OS, the fan won't
come on when it the temperature hits this trigger value. The temperature
would of course, continue to rise.
This die temperature sensor has nothing to do with the PowerBook's
fan. It's built into all PPC 750 and 7400 CPUs.
Apple could use the die sensor to control the fan, but instead they
use a sensor on the motherboard of the PowerBook, as the fan is
really intended to ensure that the ambient temperature inside the
whole PowerBook doesn't get too hot. This because there are actually
things which are much more sensitive to overheating than the CPU,
like the hard drive.
Has any code been written for LinuxPPC that will let us monitor this
sensor? I know there's a whole lm_sensors package for i386 and the like,
but is there something like this for us?
For some PowerBooks / PMU versions, I think so. We will probably get
full support for things like this when Apple puts it into Darwin/OS X.
Yes, I had heard they upgraded the PMU, but I wasn't aware of how much
they had changed it. Is it just a matter of knowing some extra bits we
can switch on and off in the PMU, or is it more than that?
Probably more than that, but I don't know anything beyond the basic
fact that the PMU (both the hardware and its control program) got the
first major upgrade in years.
Tim Seufert
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/